Be more strict about affixation functions

[clemera]

See minad/marginalia#46 (comment). It seems Emacs 28 does has problems with nil values in the results and Selectrum ignores those.

[minad]

Why did you close it? Selectrum should be more strict and not accept nil values from affixations. My Marginalia fix ensures that we are always returning strings.

[clemera]

Thanks, I wasn't looking closely but it appeared to me nil values are accepted

[clemera]

Isn't your fix handling the situation when the affixation function returns nil? So upstream Emacs affixation functions return nil values no?

[minad]

No, my fix fixes the marginalia annotation functions which may return nil. The marginalia affixation function is a small wrapper function.

[clemera]

What does happen with default completion when there the affixation returns a nil value, is there an error?

[clemera]

I haven't followed how the issue evolved on the Marginalia side, sooner or later I should install Emacs 28 do be able to check such things myself. I'm still waiting for the native compile branch to get merged into master :)

[jakanakaevangeli]

What does happen with default completion when there the affixation returns a nil value, is there an error?

Yup, very early into completion--insert-strings, string-width is used on every sub-element of each element of affix-fun's return value.

[clemera]

Okay, thanks!